https://doi.org/10.5194/amt-13-2169-2020
© Author(s) 2020. This work is distributed under the Creative Commons Attribution 4.0 License.
Intercomparison of NO 2 , O 4 , O 3 and HCHO slant column measurements by MAX-DOAS and zenith-sky UV–visible spectrometers during CINDI-2
Karin Kreher1, Michel Van Roozendael2, Francois Hendrick2, Arnoud Apituley3, Ermioni Dimitropoulou2, Udo Frieß4, Andreas Richter5, Thomas Wagner6, Johannes Lampel4,7, Nader Abuhassan8, Li Ang9, Monica Anguas10, Alkis Bais11, Nuria Benavent10, Tim Bösch5, Kristof Bognar12, Alexander Borovski13,
Ilya Bruchkouski14, Alexander Cede8,15, Ka Lok Chan16,26, Sebastian Donner6, Theano Drosoglou11, Caroline Fayt2, Henning Finkenzeller17, David Garcia-Nieto10, Clio Gielen2, Laura Gómez-Martín18, Nan Hao19, Bas Henzing20, Jay R. Herman8, Christian Hermans2, Syedul Hoque21, Hitoshi Irie21, Junli Jin22, Paul Johnston23,
Junaid Khayyam Butt24, Fahim Khokhar24, Theodore K. Koenig17, Jonas Kuhn4,6, Vinod Kumar6,25, Cheng Liu26, Jianzhong Ma22, Alexis Merlaud2, Abhishek K. Mishra25, Moritz Müller15,27, Monica Navarro-Comas18,
Mareike Ostendorf5, Andrea Pazmino28, Enno Peters5,a, Gaia Pinardi2, Manuel Pinharanda28, Ankie Piters3, Ulrich Platt4, Oleg Postylyakov13, Cristina Prados-Roman18, Olga Puentedura18, Richard Querel23,
Alfonso Saiz-Lopez10, Anja Schönhardt5, Stefan F. Schreier29, André Seyler5, Vinayak Sinha25, Elena Spinei8,30, Kimberly Strong12, Frederik Tack2, Xin Tian9, Martin Tiefengraber15,27, Jan-Lukas Tirpitz4, Jeroen van Gent2, Rainer Volkamer17, Mihalis Vrekoussis5,31,32, Shanshan Wang10,33, Zhuoru Wang34, Mark Wenig16,
Folkard Wittrock5, Pinhua H. Xie9, Jin Xu9, Margarita Yela18, Chengxin Zhang26, and Xiaoyi Zhao12,b
1BK Scientific, Mainz, Germany
2Royal Belgian Institute for Space Aeronomy, Brussels, Belgium
3Royal Netherlands Meteorological Institute, De Bilt, the Netherlands
4Institute of Environmental Physics, University of Heidelberg, Heidelberg, Germany
5Institute of Environmental Physics, University of Bremen, Bremen, Germany
6Max Planck Institute for Chemistry, Mainz, Germany
7Airyx, Eppelheim, Germany
8NASA Goddard Space Flight Center, Greenbelt, Maryland, USA
9Anhui Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Hefei, China
10Department of Atmospheric Chemistry and Climate, Institute of Physical Chemistry Rocasolano, Madrid, Spain
11Laboratory of Atmospheric Physics, Aristotle University of Thessaloniki, Thessaloniki, Greece
12Department of Physics, University of Toronto, Toronto, Canada
13A. M. Obukhov Institute of Atmospheric Physics, Russian Academy of Sciences, Moscow, Russia
14Belarusian State University, Minsk, Belarus
15LuftBlick Earth Observation Technologies, Mutters, Austria
16Meteorologisches Institut, Ludwig-Maximilians-Universität München, Munich, Germany
17Department of Chemistry & Cooperative Institute for Research on Environmental Sciences (CIRES), University of Colorado, Boulder, USA
18National Institute for Aerospace Technology (INTA), Madrid, Spain
19European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT), Darmstadt, Germany
20Netherlands Organisation for Applied Scientific Research (TNO), The Hague, the Netherlands
21Center for Environmental Remote Sensing, Chiba University, Chiba, Japan
22Meteorological Observation Center and Chinese Academy of Meteorological Science, China Meteorological Administration, Beijing, China
23National Institute of Water and Atmospheric Research, Lauder, New Zealand
24National University of Sciences and Technology, Islamabad, Pakistan
2170 K. Kreher et al.: Intercomparison of NO2, O4, O3and HCHO slant columns
25Department of Earth and Environmental Sciences, Indian Institute of Science Education and Research, Mohali, Punjab, India
26School of Earth and Space Sciences, University of Science and Technology of China, Hefei, Anhui, China
27Department of Atmospheric and Cryospheric Sciences, University of Innsbruck, Innsbruck, Austria
28Laboratoire Atmosphères, Milieux, Observations Spatiales, Université de Versailles Saint-Quentin-en-Yvelines, Centre National de la Recherche Scientifique, Guyancourt, France
29Institute of Meteorology, University of Natural Resources and Life Sciences, Vienna, Austria
30Virginia Polytechnic Institute and State University, Blacksburg, VA, USA
31Center of Marine Environmental Sciences (MARUM), University of Bremen, Bremen, Germany
32Energy, Environment and Water Research Center (EEWRC), The Cyprus Institute, Nicosia, Cyprus
33Shanghai Key Laboratory of Atmospheric Particle Pollution and Prevention, Department of Environmental Science
& Engineering, Fudan University, Shanghai, China
34Remote Sensing Technology Institute, German Aerospace Center (DLR), Oberpfaffenhofen, Germany
anow at: Institute for the Protection of Maritime Infrastructures, German Aerospace Center (DLR), Bremerhaven, Germany
bnow at: Measurement and Analysis Research Section, Environment and Climate Change Canada, Toronto, M3H 5T4, Canada Correspondence:Karin Kreher (karin.kreher@bkscientific.eu)
Received: 15 April 2019 – Discussion started: 27 May 2019
Revised: 2 December 2019 – Accepted: 16 December 2019 – Published: 6 May 2020
Abstract.In September 2016, 36 spectrometers from 24 in- stitutes measured a number of key atmospheric pollutants for a period of 17 d during the Second Cabauw Intercom- parison campaign for Nitrogen Dioxide measuring Instru- ments (CINDI-2) that took place at Cabauw, the Netherlands (51.97◦N, 4.93◦E). We report on the outcome of the for- mal semi-blind intercomparison exercise, which was held un- der the umbrella of the Network for the Detection of Atmo- spheric Composition Change (NDACC) and the European Space Agency (ESA). The three major goals of CINDI-2 were (1) to characterise and better understand the differ- ences between a large number of multi-axis differential op- tical absorption spectroscopy (MAX-DOAS) and zenith-sky DOAS instruments and analysis methods, (2) to define a ro- bust methodology for performance assessment of all partici- pating instruments, and (3) to contribute to a harmonisation of the measurement settings and retrieval methods. This, in turn, creates the capability to produce consistent high-quality ground-based data sets, which are an essential requirement to generate reliable long-term measurement time series suitable for trend analysis and satellite data validation.
The data products investigated during the semi-blind in- tercomparison are slant columns of nitrogen dioxide (NO2), the oxygen collision complex (O4) and ozone (O3) mea- sured in the UV and visible wavelength region, formaldehyde (HCHO) in the UV spectral region, and NO2in an additional (smaller) wavelength range in the visible region. The cam- paign design and implementation processes are discussed in detail including the measurement protocol, calibration pro- cedures and slant column retrieval settings. Strong emphasis was put on the careful alignment and synchronisation of the
measurement systems, resulting in a unique set of measure- ments made under highly comparable air mass conditions.
The CINDI-2 data sets were investigated using a regres- sion analysis of the slant columns measured by each instru- ment and for each of the target data products. The slope and intercept of the regression analysis respectively quantify the mean systematic bias and offset of the individual data sets against the selected reference (which is obtained from the median of either all data sets or a subset), and the rms error provides an estimate of the measurement noise or dispersion.
These three criteria are examined and for each of the param- eters and each of the data products, performance thresholds are set and applied to all the measurements. The approach presented here has been developed based on heritage from previous intercomparison exercises. It introduces a quantita- tive assessment of the consistency between all the participat- ing instruments for the MAX-DOAS and zenith-sky DOAS techniques.
1 Introduction
Passive UV–visible spectroscopy using scattered sunlight as a light source provides one of the most effective methods for routine remote sensing of atmospheric trace gases from the ground. While zenith-sky observations have been used for several decades to monitor stratospheric gases such as NO2, O3, BrO and OClO (e.g. Noxon, 1975; Platt et al., 1979;
Solomon et al., 1987; Pommereau and Goutail, 1988; Richter et al., 1999; Liley et al., 2000; Hendrick et al., 2011; Yela et al., 2017), measurements scanning the sky vertically at sev- eral elevation angles between horizon and zenith have been
established more recently. In addition to total columns, the MAX-DOAS (multi-axis differential optical absorption spec- troscopy; Hönninger et al., 2004) technique also allows the derivation of vertically resolved information on a number of tropospheric species such as NO2, formaldehyde (HCHO), BrO, glyoxal, IO, HONO and SO2 (see, e.g., Hönninger and Platt, 2002; Wittrock et al., 2004; Heckel et al., 2005;
Lee et al., 2008, 2009; Sinreich et al., 2010; Frieß et al., 2011; Hendrick et al., 2014; Prados-Roman et al., 2018) as well as aerosols (see, e.g., Wagner et al., 2004; Frieß et al., 2006; Clémer et al., 2010; Ortega et al., 2016). The num- ber of MAX-DOAS instruments used worldwide has grown considerably in recent years notably in support of satellite validation (e.g. Wang et al., 2017a; Herman et al., 2018) and for urban pollution studies (e.g. Gratsea et al., 2016;
Wang et al., 2017b), and this increase in the deployment of MAX-DOAS instrumentation for tropospheric observations, together with the diversity of the designs and operation pro- tocols, has created the need for regular formal intercompar- isons which should include as many different instruments as possible.
In 2005 and 2006, two field campaigns were held at Cabauw, the Netherlands, involving MAX-DOAS instru- ments as part of DANDELIONS (Dutch Aerosol and Nitro- gen Dioxide Experiments for vaLIdation of OMI and SCIA- MACHY). This project was dedicated to the validation of satellite NO2 measurements by the Ozone Monitoring In- strument (OMI) and SCIAMACHY (Scanning Imaging Ab- sorption SpectroMeter for Atmospheric CartographY) and aerosol measurements by OMI and the Advanced Along- Track Scanning Radiometer (AATSR) (Brinksma et al., 2008). This was followed by the first Cabauw Intercompar- ison campaign for Nitrogen Dioxide measuring Instruments (CINDI), which was organised in 2009 under the auspices of the European Space Agency (ESA), the Network for the Detection of Atmospheric Composition Change (NDACC), and the European Union (EU) FP6 Global Earth Observa- tion and MONitoring (GEOMON) project. This effort re- sulted in the first successful large-scale intercomparison of both MAX-DOAS and zenith-sky ground-based remote sen- sors of NO2 and O4 slant columns (Roscoe et al., 2010).
Data sets of NO2, aerosols and other air pollution compo- nents observed during CINDI were documented in a number of peer-reviewed articles (Piters et al., 2012; Roscoe et al., 2010; Pinardi et al., 2013; Zieger et al., 2011; Irie et al., 2011;
Frieß et al., 2016), providing an assessment of the perfor- mance of ground-based remote-sensing instruments for the observation of NO2, HCHO and aerosol. Recommendations were issued regarding the operation and calibration of the in- struments, the retrieval settings and the observation strategies for use in ground-based networks for air quality monitoring and satellite data validation. Several important findings were highlighted in view of preparing future campaigns, in par- ticular (1) the need for accurate calibration and monitoring of the elevation angle of MAX-DOAS scanners and (2), for
intercomparison purposes, the importance of synchronising measurements in time and space very accurately. The lack of such a synchronisation was indeed regarded as being respon- sible for a large part of the scatter observed during CINDI (Roscoe et al., 2010), which limited the interpretation of the results.
Seven years after CINDI, a second campaign (CINDI- 2) was undertaken at the same site (Cabauw Experimen- tal Site for Atmospheric Research – CESAR) from 25 Au- gust until 7 October 2016. Its goal was to intercompare the new and extended generation of ground-based remote- sensing and in situ air quality instruments. The interest of ESA in such intercalibration activities is motivated by the ongoing development of several UV–visible space missions targeting air quality monitoring such as the Copernicus Sen- tinel 5 Precursor (S5P) satellite launched in October 2017 and the future Copernicus Sentinel 4 and 5 satellites. The validation and ongoing support of measurements from such space missions is essential and requires dedicated ground- truth measurement systems. Because tropospheric measure- ments from space-borne nadir UV–visible sensors show little or no vertical discrimination and inherently provide measure- ments of the total tropospheric amount, surface in situ mea- surements are generally unsuitable for such a validation ef- fort. Instead, validation requires a technique that can deliver column-integrated and vertically resolved information on the key tropospheric species measured by satellite instruments such as NO2, HCHO and SO2 with a horizontal represen- tativeness compatible with the resolution of space measure- ments (e.g. 3.5 km×7 km for S5P).
Hence, the specific goals of CINDI-2 were to support the creation of high-quality ground-based data sets as needed for long-term measurements, trend analysis and satellite data validation. To achieve this, it is essential to characterise the differences between a large number of MAX-DOAS and zenith-sky DOAS instruments and analysis methods and to assess the participating instruments in their ability to retrieve the same geophysical quantities (i.e. slant columns of NO2, O4, HCHO and O3) when measured and processed in a con- trolled way (i.e. using a prescribed measurement protocol and retrieval settings). The design of CINDI-2 and the devel- opment of the measurement protocol, adhered to specifically during the official intercomparison phase, was based on the experience gained during the first CINDI in 2009 as well as more recent projects and campaigns such as the Multi-Axis Doas – Comparison campaign for Aerosols and Trace gases (MAD-CAT) in Mainz, Germany, in 2013 (e.g. Peters et al., 2017).
This paper is organised as follows. In Sect. 2, the cam- paign design is discussed including an overview of the par- ticipating groups and their instruments, and a discussion of the measurement protocol details. In Sect. 3, the results of the semi-blind slant column intercomparison are presented, and in Sect. 4, a systematic approach is proposed to quantita- tively assess the performance of the participating instruments
2172 K. Kreher et al.: Intercomparison of NO2, O4, O3and HCHO slant columns for the different target trace gas data products. Section 5 pro-
vides recommendations for observation networks and future intercomparison campaigns and Sect. 6 summarises the cam- paign outcomes.
2 Intercomparison campaign design and measurement protocol
The CESAR site was accessible for the installation of the in- struments from 25 August 2016 onwards, with the formal semi-blind intercomparison being held for 17 d from 12–
28 September 2016. Here, we concentrate on this official intercomparison phase of CINDI-2, and measurements and results are discussed for this time period only. A general de- scription of the overall campaign including a more detailed discussion of the CESAR site and all ancillary measurements can be found in Apituley et al. (2020). In short, the CESAR site at Cabauw is overall a rural site, with only a few pol- lution sources nearby, but the wider vicinity of Cabauw is densely populated, with the cities of Utrecht, Amsterdam, The Hague and Rotterdam less than 60 km away and a dense highway grid within 25 km, so that the site experiences re- curring pollution events, e.g. such as from the daily morning and afternoon rush hours.
The MAX-DOAS instruments were also complemented with a suite of in situ, profiling and mobile observations, which are described in detail by Apituley et al. (2020). In particular, a long-path DOAS measuring near surface mixing ratios of NO2and HCHO but also a range of other species such as HONO and SO2 (see, e.g., Merten et al., 2011, for a description of the technique) was operated at the CESAR site for the period of the campaign. Several mobile MAX- DOAS measurements were also made around Cabauw and between Rotterdam and Utrecht (e.g. Merlaud, 2013) in ad- dition to the static observations. NO2profiles were measured with NO2 sondes (Sluis et al., 2010) and lidar (e.g. Volten et al., 2009), as well as through in situ observations using the Cabauw meteorological tower. Extensive aerosol infor- mation was also gathered using Raman aerosol lidar and in situ samplers.
2.1 Instruments
Table 1 lists the groups and instruments that were included in the CINDI-2 semi-blind intercomparison, and an overview of the relevant instrumental details is given in Table 2. Among the 36 participating instruments, 17 were two-dimensional (2-D) MAX-DOAS systems allowing for scans in both ele- vation and azimuth, 16 were one-dimensional (1-D) MAX- DOAS systems performing elevation scans in one fixed azimuthal direction, 1 was an imaging DOAS instrument (Imaging MaPper for Atmospheric observaTions – IMPACT;
Peters et al., 2019) for which only measurements in the com- mon viewing direction were submitted, and the last 2 instru-
ments were zenith-sky DOAS systems of the SAOZ (Sys- tème d’Analyse par Observation Zénithale) (Pommereau and Goutail, 1988) and most recent Mini-SAOZ version. The complete technical specifications for each instrument can be found in Sect. S3 of the Supplement.
Instruments have been sorted into different categories.
Custom-built systems refer to instruments developed by scientific organisations for their own research activities.
Other categories denote commercial systems of various types. Pandora instruments (Herman et al., 2009) are be- ing developed at NASA/LuftBlick, commercialised by the SciGlob company and deployed as part of the Pandonia Global Network (PGN) (http://pandonia.net/, last access:
18 March 2020). EnviMes (now: SkySpec from Airyx GmbH (http://www.airyx.de, last access: 18 March 2020)) MAX- DOAS instruments (Lampel et al., 2015) have been recently commercialised based on expertise developed at the Uni- versity of Heidelberg. Mini-DOAS instruments (e.g. Hön- ninger et al., 2004; Bobrowski, 2005) are produced in Ger- many by Hoffmann GmbH (http://www.hmm.de/, last ac- cess: 18 March 2020).
No particular guidelines were given concerning the spec- tral calibration of instruments, which means that participat- ing groups were free to apply calibration steps of various levels of complexity. In addition to standard calibration pro- cedures involving dark-current and electronic offset correc- tions, wavelength registration, and slit function determina- tion, some groups performed more advanced pre-processing steps such as radiometric calibration, stray light and inter- pixel variability correction, or an explicit correction for de- tector response non-linearity, the latter being a known feature of Avantes spectrometers.
2.2 Campaign design
To allow for optimal synchronisation of the measurements, all the spectrometers participating in the semi-blind inter- comparison exercise were installed in close proximity to each other on the remote-sensing site (RSS) of the CESAR station (see Fig. 1 and Apituley et al., 2020). To achieve this, mo- bile units (similar to shipping containers) were temporarily installed for the campaign period.
The rationale behind this setup was to arrange the instru- ments in such a way as to minimise ambiguity in air masses observed simultaneously by all spectrometers. This is essen- tial for tropospheric NO2 but also for aerosol and HCHO, since all these species can feature rapidly changing concen- trations in both space and time. Considering the large num- ber of systems that needed to be accommodated, two rows of containers were deployed with the bottom row being similar to the one deployed during the previous CINDI. This bottom row of containers was predominantly used to host the 1-D MAX-DOAS instruments and the two zenith-sky systems. A second row of containers was deployed on top of the first one, with the stacked double containers providing additional
Table 1.List of participating groups and corresponding instrument IDs in alphabetical order according to their acronym.
Institute Country Acronym Instrument ID
Anhui Institute of Optics and Fine Mechanics China AIOFM aiofm-1
A. M. Obukhov Institute of Atmospheric Physics, Russian Academy of Sciences, Moscow
Russia AMOIAP amoiap-2
Aristotle University of Thessaloniki Greece AUTH auth-3
Royal Belgian Institute for Space Aeronomy Belgium BIRA-IASB bira-4
University of Natural Resources and Life Sciences, Vienna
Austria BOKU boku-6
Belarusian State University Belarus BSU bsu-5
Chiba University Japan CHIBA chiba-9
China Meteorological Administration China CMA cma-7, cma-8
Spanish National Research Council Spain CSIC csic-10
University of Colorado USA CU-Boulder cu-boulder-11, cu-boulder-12
Deutsches Zentrum für Luft- und Raumfahrt/University of Science and Technology of China
Germany/China DLR-USTC dlrustc-13, dlrustc-14
Indian Institute of Science Education and Research Mohali
India IISERM iiserm-16
National Institute for Aerospace Technology Spain INTA inta-17
University of Bremen Germany IUP-Bremen iupb-18, iupb-37
University of Heidelberg Germany IUP-Heidelberg iuph-19
Royal Netherlands Meteorological Institute The Netherlands KNMI knmi-21, knmi-22, knmi-23
Laboratoire Atmosphères, Milieux, Observations Spatiales
France LATMOS latmos-24, latmos-25
Ludwig-Maximilians-Universität München Germany LMU-MIM lmumim-35
LuftBlick Earth Observation Technologies Austria Luftblick luftblick-26, luftblick-27, luftblick-260, luftblick-270
Max Planck Institute for Chemistry, Mainz Germany MPIC mpic-28
NASA Goddard Space Flight Center USA NASA nasa-31, nasa-32
National Institute of Water and Atmospheric Research New Zealand NIWA niwa-29, niwa-30
National University of Sciences and Technology Pakistan NUST nust-33
University of Toronto Canada UTO uto-36
Figure 1.Picture of the CINDI-2 container layout at the main campaign site showing the organisation of the MAX-DOAS instruments on two superposed rows of mobile units (similar to shipping containers).
2174 K. Kreher et al.: Intercomparison of NO2, O4, O3and HCHO slant columns
Table2.Overviewofthemaincharacteristicsoftheinstrumentstakingpartinthesemi-blindintercomparisoncampaign.Thetableliststhetype,specificIDandmodelnameforeachparticipatinginstrument(columns1–3).Incolumns4–5,italsospecifieswhetherinstrumentscouldtakeazimuthalscans(ASc)and/orbeoperatedindirect-sunmode(DS).Thespectralrange,spectralresolutionandfieldofview(FOV)aresummarisedincolumns6–8.NotethattheFOVgivenincolumn8isthevalueprovidedaspartoftheinstrumentspecificationwhichmaydifferfromtheeffectiveFOVshowninFig.6.Lightcoupling(column9)denoteswhetherspectrometerswerefedbymeansofopticalfibres(F)orusingatelescopeorlensdirectlycoupledtotheentranceslit(D).Thedetectortypeisspecifiedincolumn10aseitheracharge-coupleddevice(CCD)oralineararray(LinArr),andthedetectortemperatureislistedincolumn11.
InstrumentInstrumentInstrumentAScDSSpectralSpectralFOVLightDetectorDetectorTtypeIDnamerange(nm)res.(nm)(◦)coupl.type(◦C)
Custom-builtbira-42-DMAX-DOASyy300–390/400–5600.37/0.581.0/0.5FCCD−50/−50MAX-DOASiupb-182-DMAX-DOASyn305–390/406–5790.5/0.851.0FCCD−35/−30boku-62-DMAX-DOASyn419–5530.80.8FCCD−60cu-boulder-112-DMAX-DOASyy325–470/430–6800.7/1.20.7FCCD−30cu-boulder-121-DMAX-DOASnn300–465/380–4900.8/0.50.7FCCD−30/0inta-17RASAS-IIIyn420–5400.551.0FCCD−30mpic-28TubeMAX-DOASnn305–4640.60.7FCCD20niwa-30ACTON275MAX-DOASnn290–363/400–4600.540.5FCCD−20uto-362-DMAX-DOASyy340–5600.750.62FCCD−70auth-3Phaethonyy300–4500.41.0FCCD5aiofm-12-DMAX-DOASyn290–3800.40.2FCCD−30chiba-9CHIBA-UMAX-DOASnn310–5150.4<1FCCDambientTcsic-101-DMAX-DOASnn300–5000.50.7FCCD−70amoiap-22-portDOASnn315–385/395–465/420–4900.40.3FCCD−40bsu-5MARSBnn300–5000.40.2–1.0DCCD−40iupb-37Imaging-DOASyn420–5000.81.2FCCD−30
Pandoraknmi-23Pandora-1Syy290–5300.61.5FCCD20luftblick-26Pandora-2Syy280–5400.61.5FCCD20luftblick-260Pandora-2Syy380–9001.11.5FCCD20luftblick-27Pandora-2Syy280–5400.61.5FCCD20luftblick-270Pandora-2Syy380–9001.11.5FCCD20nasa-31Pandora-1Syy280–5400.61.6FCCD20nasa-32Pandora-1Syy280–5400.61.6FCCD20
EnviMesiuph-192-DEnviMesyy300–460/440–5800.6/0.5<0.5FCCD20dlrustc-131-DEnviMesnn300–460/450–6000.6/0.60.4FCCD20dlrustc-141-DEnviMesnn300–460/450–6000.6/0.60.4FCCD20niwa-291-DEnviMesnn305–460/410–5500.6<0.5FCCD20lmumim-352-DEnviMesyn300–460/450–6000.6/0.90.4FCCD20
Mini-DOASHoffmanncma-7Mini-DOAS-UVnn300–4500.70.8FLinArrambientTGmbHcma-8Mini-DOAS-Visnn400–7101.60.8FLinArrambientTiiserm-16Mini-DOAS-UVnn316–4660.70.7FCCD−10.4knmi-21Mini-DOAS-UVnn290–4430.60.45FLinArr−5knmi-22Mini-DOAS-Visnn400–6000.50.4FLinArr−5nust-33Mini-DOAS-UVnn320–4650.71.2FCCDambientT
SAOZlatmos-24SAOZnn270–6401.320DLinArrambientTlatmos-25Mini-SAOZnn270–8200.78FCCDambientT
height. All 2-D MAX-DOAS systems were installed on the roof of the top-level containers allowing for more flexibility on the azimuth scan settings and avoiding any risk of inter- ference with the 1-D systems. All the 1-D MAX-DOAS in- struments used the same azimuth viewing direction of 287◦ (i.e. approximately WNW, with north (N) being 0◦and east (E) 90◦, etc.) which was already used during the first CINDI since it provided an unobstructed view to the horizon. This direction was also one of the azimuth directions used by the 2-D MAX-DOAS systems (see also discussion of the mea- surement protocol in Sect. 2.4).
In Sect. 2.4–2.6, further procedures aiding the compara- bility of the MAX-DOAS measurements such as the over- all measurement protocol, elevation angle calibrations and slant column retrieval settings are discussed in more de- tail. Prescribing these procedures as strictly as possible was highlighted as important during previous campaigns (see in particular Roscoe et al., 2010) and the campaign design of CINDI-2 focused on implementing such recommendations.
2.3 Semi-blind intercomparison
As in previous intercomparison campaigns of the same type (see, e.g., Vandaele et al., 2005; and Roscoe et al., 1999, 2010), a semi-blind intercomparison protocol was adopted.
The CINDI-2 exercise had three key objectives: (1) to char- acterise the differences between a large number of mea- surement systems and approaches, (2) to discuss the perfor- mance of the various types of instruments and define a robust methodology for performance assessment, and (3) to provide guidelines to further harmonise the measurement settings and analysis methods. The adopted semi-blind intercomparison protocol was based on the following approach.
a. The data acquisition schedule applied by the partici- pants was strictly prescribed to coordinate the timing and geometry of each individual measurement as ex- actly as possible, so that the same air mass could be measured by all instruments with good synchronisation.
b. For each data product, a set of retrieval settings and pa- rameters was prescribed (see Appendix A). These were mandatory for participation in the semi-blind exercise.
The data analysis software, however, was not prescribed and the different software types used by each institute are listed in Table 3.
c. All slant column data sets measured during the previ- ous day were submitted to an independent campaign referee (Karin Kreher) and her assistant (Ermioni Dim- itropoulou) every morning by 10:00 local time. At daily meetings in the afternoon (usually at 16:00), the results of the slant column comparison for measurements from the previous day were displayed anonymously, i.e. with- out any assignment to the different instruments. Basic analysis plots exploring the differences in the data sets
measured during the previous days were shown and dis- cussed.
d. The referee notified instrument representatives if there was an obvious problem with their submitted data set so that this issue could be addressed and, if possible, corrected for the remainder of the campaign.
e. After the formal campaign had finished, all participants had about 3 weeks to undertake the analysis accord- ing to the prescribed measurement and analysis protocol (see Sect. 2.4), and the final slant column data sets had to be submitted by 18 October 2016. After this date, any resubmissions were only accepted if the group could clearly state the reasons why the data set needed to be updated, e.g. if an error was found in the analysis and needed to be remedied. Further details on this process are given in Sect. 3.3 and Appendix B.
The semi-blind intercomparison exercise focused on a lim- ited number of key data products of direct relevance for satel- lite validation and NDACC operational continuity. These data products are listed in Table 4. Depending on the specific characteristics of their instrumentation, participants were free to submit all or only a subset of the data products.
2.4 Measurement protocol
As discussed above, it was recognised in previous intercom- parison campaigns (see in particular Roscoe et al., 2010) that the achievable level of agreement between MAX-DOAS sen- sors is often limited by imperfect co-location and a lack of synchronisation. This problem is especially critical for tro- pospheric NO2comparisons because of the large variability of this pollutant on very small scales. However, it is also rele- vant for other gases such as HCHO, O4, SO2and glyoxal. For this reason, it was decided to co-locate all the MAX-DOAS instruments on the same observation platform (see Sect. 2.2) and additionally to impose a strict protocol on the timing of the spectral acquisition.
The baseline for all MAX-DOAS instruments was to point towards a fixed azimuth direction (287◦) throughout the day. This direction was chosen because of the very close to obstruction-free line of sight towards the horizon. In addition, the 2-D MAX-DOAS instruments performed az- imuthal scans simultaneously according to a strict measure- ment schedule. The scheme described below was designed to ensure the maximum of synchronisation between the same type of instruments (e.g. azimuthal scans by 2-D MAX- DOAS) but also between the different types of instruments (1-D and 2-D MAX-DOAS and zenith-sky DOAS). A dis- tinction was made between twilight (morning and evening) and daytime conditions, for which separate data acquisition protocols were prescribed. According to the geometry of the solar position during the campaign, the daytime period (ex- cluding twilight) was defined to be from 06:00 to 16:45 UTC
2176 K. Kreher et al.: Intercomparison of NO2, O4, O3and HCHO slant columns Table 3.Overview of analysis software used by each of the participating institutes.
Data analysis software Institute acronym
QDOAS AUTH, BIRA-IASB, CSIC, CU-Boulder, LMU-MIM
QDOAS and WinDOAS AIOFM, NUST
QDOAS and software developed in-house UTO
DOASIS DLR-USTC, IUP-Heidelberg
DOASIS and WinDOAS IISERM,
DOASIS and software developed in-house (STRATO) NIWA
WinDOAS CMA, MPIC
WinDOAS and software developed in-house BSU
Blick Software Suite LuftBlick, NASA
Blick Software Suite and software developed in-house KNMI
NLIN BOKU, IUP-Bremen
LANA INTA
SAOZ SAM v5.9 and Mini-SAOZ software developed in-house LATMOS JM2 (Japanese MAX-DOAS profile retrieval algorithm, version 2) CHIBA Andor Solis and software developed in-house AMOIAP
Table 4.Data products included in the semi-blind intercomparison exercise and wavelength intervals selected for the analysis. Perfor- mance limits on bias (deviation from unity slope), offset and rms of dSCD linear regressions are also listed for each of the eight data products.
Data Spectral interval Bias Offset rms
product (nm) (%) (molec. cm−2) (molec. cm−2) NO2vis 425–490 5 1.5×1015 8.0×1015 NO2visSmall 411–445 5 1.5×1015 8.0×1015
NO2uv 338–370 6 2.0×1015 1.0×1016
O4vis∗ 425–490 5 0.7×1042 3.0×1042 O4uv∗ 338–370 6 0.8×1042 3.0×1042 HCHO 336.5–359 10 5.0×1015 1.0×1016
O3vis 450–520 4 0.2×1018 1.0×1018
O3uv 320–340 4 1.0×1018 4.0×1018
∗Note: the units for O4are molec.2cm−5.
with 06:00 UTC corresponding to a solar zenith angle (SZA) of approximately 83–87◦and 16:45 UTC to an SZA of ap- proximately 76–82◦, depending on the exact date during the campaign.
To allow for an NDACC-type intercomparison of strato- spheric measurements (e.g. Vandaele et al., 2005), zenith-sky twilight observations were also performed. The acquisition scheme for the dawn observations prescribed 39 measure- ments with a duration of 3 min each (integration time: 170 s;
overhead time: 10 s), starting at 04:00:00 UTC and ending at 05:57:00 UTC. This sequence was followed by a 180 s (3 min) interval allowing for a transition to the MAX-DOAS mode of which the first scans started at 06:00:00 UTC. For measurements at dusk, 40 acquisitions were recorded with a duration of 180 s each starting at 16:45:00 UTC and ending at 18:45:00 UTC.
During daytime, the acquisition scheme for MAX-DOAS and zenith-sky systems included four sequences of 15 min per hourly slot starting at 06:00:00 UTC. Individual acqui- sitions (at one given angle) were set to 1 min long in all cases. For 1-D systems, the pointing azimuth direction was set to 287◦ with elevation angles of 1, 2, 3, 4, 5, 6, 8, 15, 30 and 90◦. For 2-D systems, the azimuth angles 45, 95, 135, 195, 245 and 355◦ were successively sampled in ad- dition to the reference angle of 287◦. In each azimuthal di- rection, four elevation angles (1, 3, 5, 15◦) were scanned ex- cept for the reference azimuth of 287◦, where the same ele- vations as prescribed for the 1-D MAX-DOAS systems were used. One zenith reference spectrum was recorded every 15 min, and for 2-D systems or instruments equipped with a sun tracker, almucantar scans and/or direct-sun measure- ments were performed between the 10th and 15th minute of the sequence. For zenith-sky instruments, 1 min long acqui- sitions were performed during the whole day from 06:00:00 to 16:44:00 UTC.
Figure 2a provides an overview of the number of days each instrument was on duty during the intercomparison period. It also illustrates (Fig. 2b) the accuracy with which the different groups were able to match the imposed measurement proto- col. As can be seen, the instruments were in operation most of the time during the 17 d of the semi-blind period and most of them were able to follow the schedule to better than 1 min.
In comparison to past campaigns, the level of synchronisa- tion was clearly improved, which significantly reduced the need for smoothing or interpolating data in time. As a result, the impact of the atmospheric variability on the data com- parisons could be reduced considerably but not completely eliminated (see Sect. 3.7).
As discussed above, the measurement procedure was strict, but in spite of this comprehensive protocol, there was still some freedom left on how to implement details of the ac-
Figure 2. (a)The number of days when instruments were on duty during the 17 d intercomparison period. (b)The mean and standard deviation of the time deviations (in decimal minutes) observed in the MAX-DOAS measurements as reported by each participating group with respect to the measurement schedule defined for the campaign. Note that the instruments are listed in order of how they are categorised, and this is further explained in Sect. 4.
quisitions. For example, for managing the acquisition time, most groups decided to move the telescope and gather the spectra within the prescribed 1 min time period, while the National Institute for Aerospace Technology (INTA) (inta- 17) gathered spectra for 1 min and then moved the telescope.
As a result, a time shift was accumulated when compared to other groups (see Fig. 2). Chiba-9 also shows a notice- able time shift due to constraints in the acquisition software that prevented the strict implementation of the protocol. In the case of niwa-30, the large time shift in the UV was due to instrument-imposed alternating between measurements in the visible and UV wavelength regions (hence only one spec- tral range could be synchronised with the protocol).
Likewise, it must be noted that Pandora instruments also take separate measurements for the visible and the UV range, where a blocking filter is inserted into the optical path for the UV measurements in order to reduce spectral stray light.
Therefore, a compromise had to be found in the time syn- chronisation bracketing the requested measurement time.
This is the reason for the systematic offsets for Pandoras in Fig. 2b. Another consequence of this was that the total measurement time of Pandora instruments was about half the time of the other participating instruments, which affects the noise levels for Pandoras described throughout this paper to some extent.
2.5 Calibration of the MAX-DOAS elevation scans Because of the importance of the elevation pointing accu- racy for MAX-DOAS measurements at low elevation and as recommended after the first CINDI (Roscoe et al., 2010), dif-
ferent calibration tests involving all the participating instru- ments were undertaken during both the warm-up and semi- blind intercomparison phases. Three different approaches were used.
– On several evenings, MPIC (Max Planck Institute for Chemistry) installed an Opel car 1999 xenon lamp with a 17 cm diameter lens at a distance of 1280 m from the measurement site (angular lamp extension∼0.008◦) in the main viewing azimuth direction (287◦) of the MAX- DOAS instruments. It served as a common light source at long distance, and MAX-DOAS instruments recorded downward and upward scan spectra pointing towards the lamp.
– A white stripe on a black target at known elevation close to the instruments was scanned.
– Intensities were measured regularly during horizon scans (see Sect. 3.2 for details).
Additional calibration measurements using a near-distance lamp placed a few metres away from instruments were also performed by IUP-Heidelberg and several other groups.
Overall, these calibration procedures allowed the pointing accuracy of the different instruments and their stability dur- ing the campaign to be fully characterised (see Donner et al., 2020). As such they played an important role for the interpretation of the semi-blind intercomparison results (see Sect. 3.7).
2178 K. Kreher et al.: Intercomparison of NO2, O4, O3and HCHO slant columns
Figure 3.Hourly sunshine duration (yellow area) and temperature at the surface (red line) during the intensive campaign (topmost row), the intensity measured in the zenith and the colour index (second row from top), and the variability of the various trace gas slant column measurements performed during the semi-blind intercomparison exercise (all other rows). Slant column data measured at the main azimuth viewing direction (287◦) with the IUP-Bremen instrument (iupb-18) are shown. Green lines and symbols represent zenith-sky measurements, red lines and symbols off-axis data at 30◦elevation, and blue lines off-axis measurements up to 15◦elevation.
2.6 Slant column retrieval settings
To minimise the sources of difference between measure- ments, a set of common retrieval settings and parameters was prescribed ahead of the campaign. The use of these set- tings was mandatory for participation in the semi-blind ex- ercise. The detailed spectral retrieval settings imposed for each data product referenced in Table 1 are given in Ap- pendix A. These settings were based on the NDACC protocol for UV–vis measurements (http://www.ndaccdemo.org/data/
protocols, last access: 19 March 2020) as well as results from the first CINDI (e.g. Pinardi et al., 2013), MAD-CAT (http:
//joseba.mpch-mainz.mpg.de/mad_analysis.htm, last access:
19 March 2020) and the QA4ECV project (http://www.
qa4ecv.eu/, last access: 19 March 2020). Although not nec- essarily optimal, they represent a common baseline applica- ble to all data sets in a consistent way. Concerning the choice of the Fraunhofer reference spectrum, daily reference spectra obtained from the mean of all zenith-sky spectra acquired be-
tween 11:30:00 and 11:41:00 UTC were used. Slant columns retrieved against this reference spectrum are hereinafter re- ferred to as differential slant column densities (dSCDs).
Note that additional retrievals were also performed using sequential reference spectra (zenith-sky observations taken close to the time of the respective horizon measurements).
These data were, however, not included in the formal semi- blind intercomparison since they essentially lead to simi- lar comparison results as the analyses using daily reference spectra. They were also not available from all groups. More- over, the use of daily reference spectra presents the advan- tage of being directly applicable to twilight measurements and provides a better test of the instrumental stability over several hours of operation. As already noted in Sect. 2.1, the determination of the instrumental slit function and its even- tual wavelength dependence was the responsibility of the par- ticipating groups.
3 Semi-blind intercomparison results
3.1 Overview of slant column measurements and meteorological conditions
The meteorological conditions during CINDI-2 were excep- tionally favourable for the location and season. The upper- most row of Fig. 3 shows the hourly sunshine duration and surface temperature records for the whole semi-blind inter- comparison period (for more details, see Apituley et al., 2020). The first 4 d of the semi-blind phase were charac- terised by a clear sky with some haze in the morning and very high air temperatures for the season (>30◦C), allow- ing for efficient formaldehyde production. The next 7 d were cloudier with lower temperatures. The last 6 d of the semi- blind intercomparison exercise were also characterised by mostly clear sky or occasionally broken cloud conditions.
All other panels of Fig. 3 display the time variation in each of the dSCD data products included in the intercompar- ison, as measured by the IUP-Bremen instrument, which had excellent data coverage throughout the campaign duration.
Green lines represent zenith-sky measurements, red lines off- axis data at 30◦elevation, and blue lines off-axis measure- ments up to 15◦elevation. Results show a large variability of the NO2, O4and HCHO tropospheric columns while ozone data display the expected regular diurnal pattern mainly due to the variation in the stratospheric light path during the as- cent and descent of the sun. Due to the unusually favourable weather conditions, higher than expected values were ob- served for tropospheric HCHO while tropospheric NO2was at its lowest during the first Sunday (18 September) of the in- tercomparison campaign. The variability of the tropospheric trace gas content and the exceptionally large number of clear- sky sunny conditions were ideal for comparison purposes.
3.2 Horizon scans
Horizon scans, which consist of measuring the change in in- tensity when scanning the sky radiance across the horizon line, were systematically performed every day at noon during the semi-blind intercomparison period. Although difficult to calibrate absolutely because the horizon is generally not free of obstacles (e.g. trees, buildings or terrain height fluctua- tions), they provide a simple and valuable technique for mon- itoring the elevation pointing stability of MAX-DOAS in- struments. Figure 4 shows an example of the variation in the intensity at 440 nm, as reported by the IUP-Bremen instru- ment (blue circles). Considering that the intensity measured as a function of the elevation angle yields the integral over the telescope’s point spread function, measurements were fit- ted using an error function (Gaussian integral) according to Eq. (1)
Figure 4.Horizon scans measured by IUP-Bremen on 14 Septem- ber 2016 in the visible wavelength range. The blue circles display the intensity at 440 nm plotted as a function of the elevation an- gle reported by the instrument. Measured points are fitted by least- squares minimisation using an error function (blue line) allowing to estimate the horizon elevation (χ0) and effective field of view (FWHM) (see Sect. 3.2). The corresponding Gaussian curve (ana- lytical derivative of the fitted blue curve) is represented in red.
S=A
erf
x−x0 B
+1
+C (x−x0)+D, (1) wherexis the elevation angle andA,B,CandDare fitting parameters. The centre (x0), also fitted, provides a measure of the horizon elevation.
The analytic derivative of Eq. (1) is a Gaussian curve of which the full width at half maximum (FWHM) is given by FWHM=2p
ln(2) B. (2)
We used this quantity to estimate the effective field of view (FOV) of the instrument (see Fig. 4, red line).
Applying this fitting methodology, horizon scans delivered daily by each group were systematically analysed. Figure 5 presents an overview of the time evolution of the horizon ele- vation derived from each instrument (and their median values represented by red lines), all of them being measured in the visible wavelength range except for knmi-21. The same anal- ysis was also performed at UV wavelengths. A summary of the resulting median and 1σstandard deviation FOV derived from each instrument is presented in Fig. 6.
The time series of horizon scans provide a useful assess- ment of the stability and precision of the elevation point- ing devices used by the different instruments. In some cases, horizon scans allowed the identification of calibration biases, which could then be addressed by the instrument teams and corrected straight away. This is in particular the case for the
2180 K. Kreher et al.: Intercomparison of NO2, O4, O3and HCHO slant columns
Figure 5.Time series of horizon elevation values (blue circles) derived from daily horizon scans performed with each instrument during the intercomparison period in the visible wavelength range (except for knmi-21). When no data are available for the horizon scan analysis, a short explanation is given. The red lines indicate the median values.
dlrustc-13 and dlrustc-14 instruments. Considering the effec- tive field of view (FOV), a large variability between the in- struments was identified. This generally reflects differences in the optical design of the different systems. However, hori- zon scans can also be influenced by atmospheric conditions and by perturbations of the light intensity at the horizon (e.g.
due to fog, high aerosol loads or refraction at temperature inversions). Nevertheless, it is striking to note in Fig. 6 that horizon elevations tend to be systematically higher at visible wavelengths than at UV ones. Likewise, FOVs measured in the UV tend to be wider than in the visible range. This varia- tion is larger than expected from typical chromatic aberration effects in telescope lenses. The reason for this behaviour is not fully understood, but it is likely related to the wavelength dependence of the surface albedo, which may affect the hori- zon scan fitting process (for more details, see Donner et al., 2020).
3.3 History of slant column data set revisions
As described in Sect. 2.3, semi-blind dSCD data sets had to be submitted by 18 October 2016, i.e. 3 weeks after the end of the formal intercomparison period. However, resub- missions were accepted after this date when a clear justifica- tion was provided for the change. The main motivation for accepting late revisions was to remedy well-identified mis- takes. Details of the submitted revisions, including justifica- tions for the changes and corresponding dates, are given in Appendix B.
3.4 Pre-processing of the slant column data
Before further processing, the dSCD measurements from all groups were checked to remove unphysical values and ob- vious outliers. For this purpose, the following filters were applied: (1) dSCD data exceeding 10 times the daily me-
Figure 6.Summary of the average horizon elevation(a)and of the field of view(b), resulting from the horizon scans performed at 340 and 440 nm. Symbols represent median values and the vertical bars 10th and 90th percentiles.
dian values from the instrument were excluded, and (2) data points with a fitting rms exceeding 4 times the daily median rms were removed.
In addition, the results from the horizon scan analysis (see Sect. 3.2) were used to readjust the elevation angle of instru- ments presenting absolute elevation offsets larger than 1.5◦. This correction was performed assuming a reference horizon elevation of 0.1◦, as determined independently using lamp measurements performed at night combined with an analysis of terrain height variations (Donner et al., 2020). The im- pact of this angular correction is illustrated in Fig. 7 for NO2 dSCD measurements, which are here represented in terms of their relative difference with respect to median values from a selection of the participating instruments (for more details, see Sect. 3.5 and Fig. 8). As can be seen, the large biases observed during the first few days of the campaign for some instruments were due to systematic mispointing effects well compensated for by the correction. The impact of the cor- rection is largest for NO2, but it is also significant for other tropospheric species, in particular O4. This again stresses the importance of accurately calibrating the elevation scanner of MAX-DOAS instruments.
3.5 Determination of reference comparison data sets As in previous campaigns, the intercomparison of dSCD measurements was based on pre-selected reference data sets.
In CINDI-2, these were based on the calculation of median dSCDs obtained from a selection of measurements present- ing an acceptable agreement. Here, the selection of the refer-
ence groups, different for each data product, was performed after an initial regression analysis using the median of all data as reference. Only groups satisfying the performance crite- rion for the regression slopes were retained (see Sect. 4 and Table 4 for more details). The data sets included in the me- dian references are displayed in Fig. 8 for both MAX-DOAS and zenith-sky twilight data products. In the particular case of HCHO, the selection was performed through visual in- spection of the dSCD comparisons. Only data sets display- ing consistent behaviour at 30◦ elevation (the angle gener- ally used to retrieve first-guess total tropospheric columns using the geometrical approximation; see Hönninger and Platt, 2002) were retained for building the reference. This can be appreciated in Fig. 9 where time series of the HCHO dSCDs measured by each group are compared to the refer- ence values. As can be seen, many data sets display noisy and/or unphysical negative values and only the four selected groups (bira-4, iupb-18, mpic-28 and niwa-29) present mutu- ally consistent values. Note that a similar approach was used for the selection of the HCHO dSCD reference in Pinardi et al. (2013).
3.6 Initial assessment of the overall agreement between measurement data sets
Tables 5 and 6 show the mean relative differences (in per- cent) from the reference dSCDs and their first σ standard deviation for all participating instruments and, respectively, for all MAX-DOAS products and for all zenith-sky DOAS products. Extreme outliers (values exceeding percentile 97)
2182 K. Kreher et al.: Intercomparison of NO2, O4, O3and HCHO slant columns
Figure 7.Relative differences of NO2dSCDs (in the visible wavelength region) with respect to the median from all instruments measured during the whole semi-blind intercomparison phase for the 287◦azimuthal direction and 1◦elevation angle.(a)Results before correction for elevation offsets;(b)same results after correction for elevation offsets derived from horizon scans. Colours and symbols represent different instruments.
Figure 8.Instrument data sets selected to build the median MAX-DOAS reference(a)and zenith-sky(b)data sets. Blue marks the data sets included in the median, while grey marks the data sets not included and white the ones not available. Note that the instruments are grouped according to their specific design as custom-built, Pandora, EnviMes, Mini-DOAS or SAOZ.
Figure 9.Comparison of HCHO dSCDs retrieved by each group at 30◦elevation (red dots), and median values (black triangles). Only the four data sets (bira-4, iupb-18, mpic-28 and niwa-29) showing consistent values and a comparatively low noise level were selected for the calculation of the HCHO median.
are excluded from the analysis as well as MAX-DOAS ozone measurements since these show very small off-axis enhance- ments (see Fig. 8). Both tables provide an overall initial as- sessment of the intercomparison results indicating that for most data products (except HCHO), instruments generally agree within a few percent for the most relevant range of ele- vation angles of 1–10◦for MAX-DOAS data and for an SZA of 80–93◦for zenith-sky twilight data. One can also see that the overall agreement between instruments is better in the visible than in the UV spectral range.
For HCHO (last two columns of Table 5), the differences between the instruments are comparatively larger and, in some cases, extreme. However, restricting the analysis to the first 4 d of the measurement campaign (when the air tem- perature was warmer and the HCHO dSCDs higher) reduces discrepancies significantly, and, although a higher spread re- mains compared to any of the other products, one can con-
clude that under such favourable conditions a large number of the participating instruments provide consistent HCHO dSCD measurements. For amoiap-2, however, the instrument was operated in different modes during different time periods with some modes being more advantageous for the HCHO data analysis than others. The group found that when only HCHO data acquired during the optimal time period are used, the mean relative difference is substantially lower (approxi- mately−16 %). More details on the instrument and the dif- ferent modes are provided in Borovski et al. (2017a, b).
Based on a recent study (Spinei et al., 2020), a small bias in the HCHO dSCDs retrieved relative to a daily zenith ref- erence by the Pandora instruments (knmi-23, luftblick-26, luftblick-27, nasa-31, nasa-32) is most likely present. This bias is caused by internally emitted HCHO from the poly- oxymethylene polymer components inside the instruments.
Further details are discussed in Spinei et al. (2020).
